Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.

@firefoxwebdevs jonah, I hate to break it to you and the LLM shaped like a product manager that’s setting the agenda for your meetings, but the only time I hear about Firefox translations in any context is when Mozilla PMs try to hold it up as an example of an ethical, low-resource, useful AI feature so they can convince to be a fan of the worthless LLM shit they’re actually there to push

the reason why I don’t hear about translations otherwise is simple: it’s shit

@zzt @firefoxwebdevs OK, now make the same argument for the spell-checker, sync, and the set of CAs, etc. etc. supplied with the browser. Its as if y'all were trained by Microsoft PR to take the arguments Mozilla used against tying IE to Windows and extend them ad-absurd-um to features in Mozilla's own browser ("just turn it around back in their faces" said the Armani suit).

Meanwhile, Red Hat is quietly undermining any legal basis for copyleft and leaning into the idea that gratis products (Fedora) shouldn't have robust & transparent system update tools. Oh and the umpteen other for-profit controlled (opposite of Mozilla) FOSS projects that get plugged in these spaces pretty much constantly. Linux Foundation being controlled by Microsoft and Google...? crickets chirping.

This is what makes me tired of IT and geek culture. Its become like everything else, just kneejerk crap with zero reflection and sense of proportion. As I hinted above, it morphs into this shadow of corporate PR. Consider, if people spent their time criticizing actual badness in Firefox, like ad tracking and DoH, that would be inconvenient for certain interests from Brave on up to Apple and Google. I think the style and quality of venting we usually see about Mozilla serves those interests, much of it probably fed by sock puppets.

@firefoxwebdevs Because the term "AI" has been so heavily overloaded to include ML, LLMs, Uncle Tom Cobly and all, including the translations in the "AI" kill switch would be signalling to users that their consent is being taken seriously - especially the way that unwanted "AI" is being included so conspicuously in so many tech products at the moment. Ask for consent, don't end up begging for forgiveness on what you see as a technicality.

@firefoxwebdevs an important addendum regarding Firefox translate: by my math (N = my replies), 25% of its users are fucking unhinged

I told them not to use the necronomicon to train the base model but here we fucking are

@zzt @firefoxwebdevs i'll have you know i'm at least 75%* hinged

* vibe estimate, but we carefully graphed it so it's data now

Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.

They're not LLMs. They're trained on open data.

Should translation be disabled if the AI 'kill switch' is active?

Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.

They're not LLMs. They're trained on open data.

Should translation be disabled if the AI 'kill switch' is active?

@davidgerard @firefoxwebdevs I know you, you’ve read much worse things than the necronomicon

you’ve written wiki articles about most of them

@firefoxwebdevs

I think the challenge with everything going on here is one of clarity.

@sil, you are asking them about disclosure of models and sourcing. But that is far from the only AI that is in the system.

The tool that does grammar checking and language identification does not leverage an LLM, and while there may be some type of model underneath, the context is very different. Tools that detect spam pages or faulty JavaScript that locks the pages, that's another type of AI hard at work.

Is the browser allowed to support speech to text?

@jmax You're calling out that Firefox may not be able to do this, but I think that mischaracterizes the scope of what's happening here.

The browser has several types of non-deterministic, probabilistic tools in it that provide useful services. Now there's a backlash against one very specific version of those non-deterministic, probabilistic tools. But the backlash is vociferous, often unsolvable, and incredibly broad.

It's hard to engage with non-specific anger.

@cassidy @firefoxwebdevs The term "AI" has existed since 1956 so of course it's going to have a very broad definition.

Things don't just stop being "AI" when AI researchers invent newer "more AI" stuff.

@twifkak Wouldn't that be a valid working definition of "open"?

Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.

They're not LLMs. They're trained on open data.

Should translation be disabled if the AI 'kill switch' is active?

@tasket It would be, much in the way that "guaranteed not to turn pink in the can" is a valid description of bad salmon [1]. A disingenuous mislead from what people really care about in the product.

[1] I know it didn't happen. It's a good metaphor.

Firefox uses on-device downloaded-on-demand ML models for privacy-preserving translation.

They're not LLMs. They're trained on open data.

Should translation be disabled if the AI 'kill switch' is active?

@zzt @pixel @firefoxwebdevs "Nobody likes our product any longer, but at least we never had to entertain any *shudder* critical feedback."

@firefoxwebdevs an important addendum regarding Firefox translate: by my math (N = my replies), 25% of its users are fucking unhinged

I told them not to use the necronomicon to train the base model but here we fucking are

@twifkak I think you're mixing up "We do not own" with "We do not have rights to".

@wes @firefoxwebdevs Sure. But can we agree that it does not represent a core functionality of a web browser?

Like "this meeting could've been an email," but "this feature could've been an add-on."

A web browser should load web pages, allow you to interact with them, and offer add-on support for functionality that doesn't match the definition of "web browser." It's all pretty straight-forward if you're not a marketer, whose brains are all broken.

@firefoxwebdevs My closest answer would be "no", but I think the question is kind of mis-phrased here, and that's probably going to lead to a confusing and potentially misleading outcome.

The problem that people have is not with "AI" as a generalized category, but with the current generation of thieving, climate-destroying, grifting systems that are marketed as AI to an overwhelming degree - notably LLMs and "generative AI", but really anything with those inconsiderate properties.

If your kill switch is presented as an "AI kill switch", then depending on the person they're either going to understand that as "exploitative tech", or as "machine learning", and so make different assumptions as to whether local translation is included in that.

So I think you'll have to be a lot more explicit about what you mean; either by describing clearly what the kill-switch includes, or what it excludes, right in the place where the option is offered. Otherwise it's damned if you do, damned if you don't; depending on whether you include translations, either one or another group is going to be upset with the unexpected behaviour.

So, ethically, if the translation feature is built on ethically collected data, and it has no outsized climate impact, then I would not consider it something that needs to be included in a "get rid of all of it" kill switch. But to convey this clearly to users, both that and why it isn't included should be explained right there with the button, with potentially a second-step option to disable it anyway if someone still feels uncomfortable with it.

That way you've transparently communicated to users and shown that you have nothing up your sleeve by immediately and proactively offering them an option to disable that, too, if they have already shown interest in removing "AI" features.

@tasket Perhaps. Show me what rights they have to it.